Source-linked AI summary
Dissecting a Social Botnet: Growth, Content and Influence in Twitter
Norah Abokhodair, Daisy Yoo, David W. McDonald
TL;DR
The paper examines how one Twitter social botnet grows, differs from regular users, and may influence relevant discussions. It finds findings that sometimes contradict and sometimes support current knowledge, while identifying distinctive content and influence tactics.
Problem
The paper addresses how a specific Twitter social botnet grows, differs in content from regular users, and may influence relevant discussions.
Method
The study examines one Twitter botnet by comparing its content with regular users and analyzing retweeting and attention-direction tactics.
Results
The botnet used misdirection and smoke screening, differed in content from regular users, and received attention from real-user accounts.
Takeaways & Limitations
The findings sometimes contradict and sometimes support current knowledge on social botnets, while characterizing bot–human interactions on Twitter.
Takeaways & Limitations
The study is limited to one specific botnet, and wider detection could include real users who probably would not recognize the botnet.
Abstract
from arXiv · showhide
Social botnets have become an important phenomenon on social media. There are many ways in which social bots can disrupt or influence online discourse, such as, spam hashtags, scam twitter users, and astroturfing. In this paper we considered one specific social botnet in Twitter to understand how it grows over time, how the content of tweets by the social botnet differ from regular users in the same dataset, and lastly, how the social botnet may have influenced the relevant discussions. Our analysis is based on a qualitative coding for approximately 3000 tweets in Arabic and English from the Syrian social bot that was active for 35 weeks on Twitter before it was shutdown. We find that the growth, behavior and content of this particular botnet did not specifically align with common conceptions of botnets. Further we identify interesting aspects of the botnet that distinguish it from regular users.
1.1. Opinion.
Opinion tweets express personal or collective reactions to the Syrian crisis, including experience, patriotism, criticism, humor, prayers, and artistic expression. Tweets are classified as News instead when linked material is identifiable as a news article.
- Opinion tweets include personal analysis, patriotism, criticism, rhetorical questions, jokes, prayers, and poems about the Syrian crisis.
- Tweets are coded as News when linked content is identifiable as a news article rather than self-expression.
- Personal-experience remarks may reference family, friends, acquaintances, or amateur videos from the war.
- Conversational tweets are identified by dialogic context, especially when removing an initial @username leaves the tweet otherwise uncodeable.
- Breaking-news tweets use urgency headings or come from accounts labeled as breaking-news aggregators.
1.5. News.
News tweets report information about the Syrian civil war from an objective or third-party stance. The category includes verified news and aggregator accounts, NGO reports, martyr notices, and reliable linked analysis or editorials.
- News tweets present Syrian civil-war information from an objective, third-party, or clearly objective-toned stance.
- The category includes reports from verified news accounts, news aggregators, NGOs, and martyr notices.
- Linked articles from reliable sources are included when they provide news analysis or editorials.
1.6. Mobilization of Resistance/Support
Mobilization of Resistance/Support covers tweets that organize collective action related to the Syrian civil war, including online petitioning and requests for retweets.
- Tweets in this category call for organizing, meetings, protests, or gatherings related to the civil war.
- The category also includes online petitions, retweet requests, and related forms of slactivism.
1.7. Mobilization for Assistance
Mobilization for Assistance covers tweets that recruit people or organizations to provide social and humanitarian aid during the Syrian civil war.
- Tweets mobilize people and organizations to provide social or humanitarian aid related to the civil war.
- Examples include requests for donations of food, clothing, or money.
- Table 1 compares tweet-content categories across the SSB, regular Arabic users, and regular English users.
1.8. Solicitation for Information
This category covers tweets asking about the situation in Syria, while rhetorical questions about the civil war are coded as “Opinion”.
- Tweets asking about the situation in Syria are classified under solicitation for information.
1.9. Information Provisioning
This category covers tweets providing information that is individually actionable.
- Information provisioning consists of tweets that provide individually actionable information.
1.10. Pop Culture.
This category covers tweets referencing celebrities, music, sports, entertainment, and related people.
- Pop-culture tweets reference celebrities, music, sports, entertainment, or related people.
1.11. Other
This category covers tweets related to Syria and other countries that are not clearly associated with the civil war.
- Other tweets concern Syria and other countries without a clear association with the civil war.
1.12. Spam/Phishing
Some tweets were classified as spam or phishing content rather than interpretable Syria-related material.
- Spam/phishing tweets were identified as a distinct content category.These tweets appeared to constitute spam or phishing activity.
- The category covered tweets that looked like spam.
- The category also covered tweets that looked like phishing.
1.13. Uncodeable
Tweets were marked Uncodeable when their language or brevity prevented reasonable interpretation and coding.
- Tweets in languages other than Arabic or English were classified as Uncodeable.
- Tweets that were too short to reasonably interpret were also classified as Uncodeable.
- Secondary codes: Secondary codes were developed independently from the primary codes and based on context.
2.1. Local Context.
Local Context covered tweets mentioning only the current situation in Syria without further specification.
- Local Context included tweets about key events in Syria.
- The category included tweets about elections in Syria.
- The category also included tweets about military reactions in Syria.
2.2. International Context.
The SSB’s content differed from regular Arabic and English users in news, opinion, spam/phishing, and off-topic material. Its high-volume retweeting sometimes reached humans and could redirect attention away from the Syrian conflict.
- Content differences: Table 1 compares content differences among SSB, regular Arabic, and regular English tweets.The English sample also contained many Uncodeable tweets because some apparent Syrian city terms appeared in other languages.
- News reporting: More than half of SSB tweets were News, compared with slightly over one-third among regular Arabic and English users.The analysis excluded retweets, so these tweets represented or described news stories rather than simply repeating them.
- Opinion: Opinion comprised about 45% of Arabic tweets, 25.8% of English tweets, and 12.4% of SSB tweets.Arabic users frequently expressed personal reactions and shared prayers, while the SSB showed comparatively little opinion content.
- Spam/phishing: About 25% of English tweets contained spam or phishing content tied to attention around the Syrian civil war.
- Other: The SSB’s Other category comprised 31.8% of its content, with International context at 59% and Local context at 41%.Many such tweets concerned topics associated with Syria but not clearly related to the conflict.
- Misdirection: Irrelevant SSB tweets could flood #Syria with unrelated topics, functioning as a smoke screen that distracted attention from the civil war.Examples included employment-program news and an article about Jimmy Savile on a pro-regime website.